Skip to content
Clear all

I did not see this one coming! (Ultimate USB-C mic improvement)

1 Posts
1 Users
0 Reactions
Posts: 476
Topic starter

What goes from 0.8% THD to 14% THD over its predicted dynamic range in a simulated environment.

The answer to this question is shocking and it went right over my head until I started work converting my own designs to suit the original DIY Perks USB-C mic.

This is one of those things I miss but the issue is with the way the original design uses a split supply to the FET. Note that the "professional" lads are working with +48V all over the place and lower voltage versions use 12-15 volts but with both supplies positive. There's no immediately obvious reason (to me) that those early designs wouldn't operate with split rails - that precedes transistors  by decades.

The answer is in the peculiar way that JFETs operate and I'm not going to dive into this right now because it's... kinda weird.

Here's a 100mV sine wave. A pure wave entering the JFET gate (in red) vs. what comes out at the source and drain:


Seems OK right? Now look at this one:


So what's happened? Well we've lost about 50% of the signal but that's no biggie, we've got a big amp on the other end and sure enough we do.

But that's not the whole story. Let's take a look at each of those signals closely.



and Source:


This isn't that clear unless you spend days looking at sine wave in simulators and on measuring gear. This represents that horrendous distortion due to non-linearity in the behaviour of JFETs.

Thankfully there is a fix even IF you've made up everything as originally specified:

Take the source pin from your MICROPHONE capsule or whatever you connected -15 to and connect that to 0V at the pre-amp.


Or replace the 2K2 resistor from the -15V (source) line to the mic with a link (green line is the new signal at the source).



That's it. You don't have to do anything with the rest of the circuit. I love it when things just work out as quick and easy as that.

Either way, we're change how current flows through the FET and keep it inside the limited area where it's still relatively linear.

Even at low levels, the current design still exhibits considerably more distortion than the modified one the main difference is that's it not that noticeable a very low levels and high-sensitivity mics like the JLI2555 (which is excellent) tend to mask that. 

Here's the test setup in LTSpice for comparison. You can get the distortion figures by pressing ALT-R (run the simulation) and then CTRL-L (log) to see the Fourier results.

This topic was modified 1 month ago by marcdraco

Take everything I say with a pinch of salt, I might be wrong!

Posted : 11/05/2024 1:17 pm